Large interannual CO2 and energy exchange variability in a freshwater marsh under consistent...

Large interannual CO2 and energy exchange variability in a

freshwater marsh under consistent environmental conditions

Adrian V. Rocha1,2 and Michael L. Goulden1

Received 10 February 2008; revised 27 June 2008; accepted 9 September 2008; published 26 November 2008.

[1] We analyzed a 5-year record of the CO2 and energy exchange, Aboveground NetPrimary Production (ANPP), maximum Leaf Area Index (LAImax), and EnhancedVegetation Index (EVI) for a Typha marsh in Southern California. The marsh was a netsource of carbon over the study, despite high rates of ANPP. Interannual Net EcosystemProduction (NEP) variability was the largest that has been reported for any terrestrialecosystem and was attributed to changes in maximum photosynthetic rates (GEEmax). Thevariation in energy and mass exchange was coupled between years; years with higher thanaverage rates of carbon uptake were associated with lower than average sensible heatfluxes. Remotely sensed measures of surface greenness (EVI) were closely related toGEEmax variation, providing further evidence of interannual variability. We were unable toattribute the fluctuations in GEEmax to the direct effects of weather on ecosystemphysiology, or to interannual variation in LAImax. GEE did not vary systematically with airtemperature or the presence of standing water in the marsh; GEEmax did not vary withLAImax between years. Rather, interannual variation in carbon exchange at the SJFMresulted from shifts in the marsh’s production efficiency (the rates of gross or net CO2

exchange per LAI) that were not caused by changes in the weather. Our findings challengethe assumptions that interannual variation of land-atmosphere exchange is universallycaused by the direct effect of weather on ecosystem physiology, and that an ecosystem’sphysiological response to the physical environment is consistent from year-to-year.

Citation: Rocha, A. V., and M. L. Goulden (2008), Large interannual CO2 and energy exchange variability in a freshwater marsh

under consistent environmental conditions, J. Geophys. Res., 113, G04019, doi:10.1029/2008JG000712.

1. Introduction

[2] Long-term records of whole ecosystem CO2, waterand energy exchanges have provided valuable insight intothe interannual dynamics of ecosystem function. Theserecords have demonstrated that exchanges of energy andmass are tightly coupled [Law et al., 2002] and thatecosystems are dynamic in their ability to capture and storecarbon from year to year [cf. Goulden et al., 1996; Chen etal., 1999; Barford et al., 2001; Flanagan et al., 2002; Dunnet al., 2007]. Most work has attributed interannual energyand mass exchange variability solely to weather. However,statistical attribution of interannual variability in Net Eco-system Exchange of CO2 (NEE) to climatic factors has beendifficult [Hollinger et al., 2004] and attempts to modelinterannual NEE variability using process based ecosystemmodels have achieved moderate to low success [Joiner etal., 1999; Kelly et al., 2000; Griffis and Rouse, 2001;Thornton et al., 2002; Hanson et al., 2004; Urbanski etal., 2007].

[3] Our inability to model and attribute interannual eco-system variability may reflect the simplicity with which weview and represent ecosystems in models. Most modelsassume that ecosystems respond to the environment pas-sively and that this response is constant over time [Hansonet al., 2004; Urbanski et al., 2007]. Most models are highlyparameterized such that ecosystem respiration is representedas a simple Q10 response with temperature, and photosyn-thesis is represented with a constant light use efficiency andcarboxylation capacity [Hanson et al., 2004]. These modelparameterizations do not allow for temporal changes inbiological factors associated with interannual changes inlight use efficiency, carbon allocation, or substrate inducedrespiration [Davidson et al., 2006]. However, interannualecosystem variability almost certainly involves internalbiological controls, which can change the response of anecosystem to the environment [Hui et al., 2003; Richardsonet al., 2007]. The core question, therefore, involves therelative importance of these internal biological dynamics incontrolling interannual variability.[4] In this study, we describe the interannual dynamics of

CO2 and energy exchanges, Aboveground Net PrimaryProduction (ANPP), maximum Leaf Area Index (LAImax),and remotely sensed vegetation indices at the Typha-dom-inated San Joaquin Freshwater Marsh (SJFM) located inSouthern California. The SJFM is a freshwater marshlocated in a maritime Mediterranean climate. Interannual

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, G04019, doi:10.1029/2008JG000712, 2008

1Department of Earth System Science, University of California, Irvine,California USA.

2Now at Marine Biological Laboratory, Woods Hole, Massachusetts,USA.

Copyright 2008 by the American Geophysical Union.0148-0227/08/2008JG000712

G04019 1 of 12

weather variability at the SJFM is small and water input tothe SJFM is managed and consistent from year to year.Typha wetlands are known for their high productivity[Bradbury and Grace, 1983] and abundance of nutrientresources compared to other terrestrial systems [LeBauerand Treseder, 2008]. We hypothesized that interannualvariability in energy and mass exchanges would be small,and that the high productivity of the SJFM would result in astrong carbon sink. We address these hypotheses by doc-umenting a long-term record of whole ecosystem CO2 andenergy exchange, assessing the ability of the SJFM tosequester carbon, determining the relationships betweeninterannual weather and ecosystem function, and comparingthe interannual SJFM ecosystem function variability to thatreported for other ecosystem types.

2. Methods

2.1. Site Description

[5] The study was conducted at the San Joaquin Fresh-water Marsh Reserve (SJFM) located on the University ofCalifornia’s Irvine campus [Goulden et al., 2007]. TheSJFM is in coastal Orange County, 3 m above sea leveland 8 km northeast of the Pacific Ocean (33� 390 44.400N,117� 510 6.100W). The climate is Mediterranean with a meanannual temperature (1975–2005) of 17.7�C and a meanannual precipitation of 350 mm. The SJFM is dominated byCattail (Typha latifolia L.) with a minor component (�20%percent cover) of Bullrush (Scirpus californicus C.A. Mey.Steud.) and willows (Salix spp.) around the periphery. Thehydrology of the SJFM has been altered and water input isnow managed. The SJFM is flooded to a depth of �1 mabove the mineral soil in the winter of most years. Naturalprecipitation is the only additional source of water afterinitial flooding and the SJFM dries by evapotranspirationand subsurface drainage through the spring and summer.Standing water disappears by midsummer in most years.

2.2. Micrometeorological Observations

[6] Meteorological data and the NEE of CO2 werecalculated at half hourly intervals from December 1998 toDecember 2003 [Goulden et al., 2007] (http://www.ess.uci.edu/�marsh). The eddy covariance tower was mountedon a floating platform that was unaffected by lateral orvertical wind movement and remained level throughout theyear. Instruments at 5.5 m above the platform surfacerecorded the environmental conditions at 0.5 Hz. Radiationmeasurements included incoming and reflected Photosyn-thetic Active Radiation (PARi and PARr) measured withquantum sensors (Model LI-190, LI-COR Inc., Lincoln,NE) and incoming and reflected shortwave radiation (Ki andKr) measured with thermopile pyranometers (Kipp & ZonenCM3, Delft, Netherlands). Vapor pressure deficit (VPD)was calculated from the measured water vapor concentra-tion and air temperature, using the Clausius-Clapeyronequation to derive the saturation vapor pressue. MiddayPAR albedo (PARalbedo) was calculated by dividing middayvalues of PARi by PARr. Air temperature, precipitation, andhumidity were recorded at the tower, but we opted to use amore continuous data set from the nearby Santa Ana JohnWayne Airport (2 km to the north; data from the U.S.National Climatic Data Center, http://www.ncdc.noaa.gov,

see Goulden et al. [2007] for further details). The JohnWayne Airport data set contained hourly air temperature,precipitation and relative humidity for 1999–2003.[7] The NEE of CO2 was calculated as the 30-min

covariance of the vertical wind velocity measured with asonic anemometer (CSAT3, Campbell Scientific, Logan,UT) and the CO2 mixing ratio measured with a closed pathinfrared gas analyzer (IRGA: Model 7000 or 6262, LI-CORInc., Lincoln, NE). Adequate turbulent mixing and energybudget closure are key factors determining the quality ofNEE data [Goulden et al., 1997]. Adequate turbulent mix-ing at the SJFM, as determined from plots of NEE versusfriction velocity (u*), occurred for u* > 0.20 m s�1. Theeddy covariance technique samples the flux from a fewhectare patch of vegetation that is upwind of the tower.Daytime winds at the SJFM almost always originate fromthe southwest, and we located the tower near the easternside of the SJFM, in an area where Typha marsh extended500- to 900-m to the southwest. The marsh extended lessthan 100 m to the northeast of the tower, and the eddycovariance measurements during periods with wind fromthis direction did not provide a valid measure of marshexchange. NEE data during periods with either wind from315� to 145� or u* < 0.20 m s�1 were excluded from theanalysis. We referred to the remaining flux data as ‘‘un-filled’’ tower data.

2.3. Separating NEE Into Component Fluxesand Gap Filling

[8] NEE is the sum of three component fluxes: GrossEcosystem Exchange (GEE), autotrophic respiration (Ra),and heterotrophic respiration (Rh). Eddy covariance deter-mines NEE, which can be partitioned into GEE and TotalEcosystem Respiration (TER) by separately considering theday and night observations [Goulden et al., 1997]. Thefurther partitioning of TER into Rh and Ra is difficult andcannot be done accurately from eddy covariance data alone:

NEE ¼ Ra þ Rhð Þ þ GEE ¼ TERþ GEE ð1Þ

Eddy covariance observations are commonly reported usingthe meteorological sign convention, where a downwardmovement of CO2 is considered a negative flux. Themeteorological GEE always has a negative sign. A morenegative meteorological GEE indicates an increase inphotosynthesis and a more positive TER indicates anincrease in respiration. NEE is negative when photosynth-esis is greater than respiration and positive when photo-synthesis is less than respiration.[9] At night, GEE is zero and NEE (NEEnight) is equal to

total ecosystem respiration. The difference between NEEday

and NEEnight can be used to calculate GEE provided that arealistic approach is adopted for extrapolating nocturnalrespiration (NEEnight) to daytime periods. Many researchershave used an exponential fit with temperature to extrapolateNEEnight to daytime conditions (GEEexponential TER) [cf.Goulden et al., 1997]. However, this approach is notwithout its weaknesses, because the respiration model isparameterized with cool nocturnal temperatures and extrap-olated to warmer daytime temperatures, and because theresult often depends on whether air or soil temperature isused [Richardson et al., 2006]. An alternative, more con-

G04019 ROCHA AND GOULDEN: LARGE INTERANNUAL VARIABILITY

2 of 12

G04019

servative approach is to assume that TER remains constantfrom day to night (GEEconstant TER).[10] We tested the sensitivity of interannual variations in

GEE, NEE, and TER to gap filling assumptions using bothapproaches:

GEEexponential TER ¼ NEE� a* exp b*Tð Þð Þ ð2Þ

GEEconstant TER ¼ NEE� NEEnight

� �ð3Þ

Rates of respiration in equation (2) were determined from astatistically derived Q10 response of NEEnight to airtemperature (T) for the entire year. We used a long timeperiod to derive Q10 respiration functions because Q10

functions based on monthly to seasonal time scales had littleto no predictive value. Rates of respiration in equation (3)were determined from unfilled averaged NEEnight from theprior 25 days. Our analysis found that the magnitudes of thefluxes changed, but the patterns of interannual variabilityremained similar among gap filling techniques (data notshown). On average, -GEE was 17% higher and TER was16% higher for the annual Q10 gap-filling approach ascompared to the simple gap-filling approach. From thisanalysis, we conclude that the Q10 gap-filling approach willoverestimate TER and GEE because temperature poorlypredicted NEEnight and errors in the exponential modelincrease with increases in temperature. Nonetheless, weemphasize that our conclusions are independent of gapfilling technique. We believe that the simple gap fillingtechnique (equation (3)) provides a more conservative androbust estimate of the SJFM annual carbon budget, and wetherefore used this approach for our subsequent analyses.[11] Missing data due to sensor malfunction, periods of

low turbulence, and inappropriate wind direction were filledusing empirical relationships to calculate the daily andannual sums. Twenty five percent of the data had to befilled for 1999–2003 period. GEE was filled using aMichaelis-Menten hyperbolic regression with Ki [Ruimy etal., 1995]. Fitted coefficients for the light curve describedthe potential maximum GEE as Ki approaches infinity, andthe quantum efficiency of GEE. A combination of theMichaelis-Menten fit and an estimate of leaf phenologyderived from a statistical relationship between NEE andreflected radiation during the 2000 growing season wasused to fill a two and a half month gap in the early 2000growing season. Respiration data were filled using averagedNEEnight from the prior 25 days. Missing Ki data were filledusing a lookup table that listed half hourly average Ki foreach period during the entire Ki record.[12] An analysis of the energy budget closure at the SJFM

indicated that the raw turbulent energy flux measurementsunderestimated the true energy flux by �20% [Goulden etal., 2007]. This percentage of unaccounted flux is similar tothat observed in many other eddy covariance studies, and ispresumably caused by transport in low-frequency circula-tions that are underestimated by a 30-min averaging interval[Twine et al., 2000]. We therefore forced our energy budgetto close by multiplying latent and sensible heat fluxes bycorrection coefficients so that the long-term sum of turbu-

lent fluxes equaled net radiation [Twine et al., 2000;Goulden et al., 2007].

3. Biomass Harvests

[13] Cattails senesce and produce aboveground materialannually, and so measurements of peak aboveground bio-mass were used as a measure of Aboveground Net PrimaryProduction (ANPP). We sampled all living plants withinthirty 0.25-m2 quadrats along a 91.5 m transect that radiatedto the southwest of the eddy covariance tower during themonth of peak biomass (September) from 1999 to 2003.Plants were pulled from the ground, clipped below thecrown to remove rhizomes and roots, taken back to thelab, oven-dried at 65�C for 2 to 3 days and weighed. Wepartitioned growth into leaves, stems, flowers, and thecrown base by sorting a subsample of plants.[14] Collected leaf material was also used to determine

the peak Leaf Area Index (LAImax) from measurements ofspecific leaf area. A subset of 25 to 35 plants from each yearwere weighed, scanned at 150 dpi using a Microtek Scan-Maker i900 (Microtek USA, Carson, Calif.), saved as jpegimages and imported into ENVI 4.0 (ITT Visual Informa-tion Solutions; Boulder, CO, USA) for determination of leafarea. The determination of leaf area in ENVI was based onpixel counts on the red band that had a threshold luminosityvalue of greater than 100. This method was tested foraccuracy against materials of known area and estimateswere found to be in excellent agreement with actual area(r2: 1.0; y = 1.0048*x; Mean Absolute Error: 0.80 cm2).Specific leaf areas were then scaled to the canopy basedon the leaf biomass measurements.[15] We obtained 5 midsummer (end of June–mid-

August) Landsat images (Landsat 5 or 7) that included theSJFM (Path: 40 Row: 37) during 1999–2003. Landsatscenes have a �30 m spatial resolution and were purchasedfrom the MultiResolution Land Characteristics Consortium(MRLC; http://www.mrlc.gov/). Landsat data were takenafter the hydroperiod, and surface water did not affect thecalculation of vegetation indices.[16] Landsat images were cross-calibrated and surface

reflectance data from a 90 by 90 m area located immediatelyupwind of the eddy covariance tower was used to calculatethe Enhanced Vegetation Index (EVI). Images were correctedfor sensor differences and drift by transforming pixel valuesinto exoatmospheric radiances [Chander and Markham,2003]. Images were then intercalibrated to remove anyatmospheric influences and transformed into surface reflec-tance using the temporally invariant surface target methodagainst a surface reflectance image that was processed by theLEDAPS project (http://ledaps.nascom.nasa.gov/ledaps/ledaps_NorthAmerica.html) [Hall et al., 1991]. Surfacereflectance images were then used to calculate the EnhancedVegetation Index (EVI) [Huete et al., 2002]. Images wereprocessed with ENVI.

4. Results

4.1. Intra-annual and Interannual Variability inEnvironmental Conditions

[17] Seasonal variations in environmental conditions atthe SJFM reflected the maritime Mediterranean climate


3 of 12

G04019

(Figure 1). Incoming solar radiation was low in the winterand peaked about a month before the peak in air temperature(Figure 1a). Over the course of a year, daily average airtemperature ranged from 8 to 22�C (Figure 1b). The vapor

pressure deficit (VPD) usually peaked about a month afterthe peak in air temperature and was lowest in the winter(Figure 1c). The SJFM was flooded to a depth of �1 mbeginning in December, January, or March of each year(Figure 1 and Table 1). The earliest arrival of water came inlate December of 2002, and the latest arrival of water camein March of 1999. The SJFM received water in late Januaryor early February in the other years.[18] Interannual variability in environmental conditions at

the SJFM from 1999 to 2003 was small (Table 1). Averageannual air temperature from 1999 to 2003 (17.4�C) wassimilar to the 30-year average (17.7�C) and ranged 1.3�C.Average minimum air temperatures ranged 1.7�C, whileaverage maximum air temperatures ranged 1.5�C. Theinterannual range in average temperature was somewhatless than that observed for a temperate forest (1.6�C [Goughet al., 2008]), a tallgrass prairie (1.6�C [Suyker et al.,2003]), and a tropical peat swamp (1.6�C [Hirano et al.,2007]), and much less than that observed for a boreal forest(3.2�C [Dunn et al., 2007]). Annually averaged VPD rangedfrom 0.53 to 0.64 kPa and averaged 0.56 kPa, whileannually integrated incoming solar radiation varied 8%and averaged 6647 MJ m�2 a�1. The hydroperiod rangedfrom 87 to 208 days.

4.2. Interannual Variability in Canopy Properties

[19] There was significant interannual variability in plantcanopy properties (Figure 2a). Maximum Leaf Area Index(LAImax) was higher in 2000, 2001 and 2003 and lower in1999 and 2002 (ANOVA; P < 0.001). There were nosignificant differences in stand density (ANOVA; p =0.21) or specific leaf area (ANOVA; p = 0.18) betweenyears, implying that plants were larger (i.e., more leaves ortaller) in 2000, 2001 and 2003. We tested this hypothesisand found that average individual plant weight was reducedin 2002 (ANOVA; P < 0.001). Rates of ANPP were similarto other productive wetlands and crops (Figure 2b) [Esser etal., 2000]. ANPP was highest in 2001 and lowest in 1999 andranged from 580 gC m�2 a�1 in 1999 to 1018 gC m�2 a�1 in2001. Interannual ANPP variability was not significantlyrelated to annual air temperature (p = 0.17), solar radiation(p = 0.44), average VPD (p = 0.78), or the length of thehydroperiod (p = 0.55).

4.3. Intra-annual and Interannual Variability inSurface Fluxes

[20] Half-hour fluxes of energy and mass at the SJFMdisplayed marked intra and interannual variability (Figure 3).Latent and sensible heat fluxes were closely tied to theintensity of sunlight and canopy phenology over the season.Sensible heat fluxes were lowest from October to Decemberand were often negative during winter nights and early

Figure 1. Daily environmental conditions at the SJFMfrom 1999 to 2003. Daily integrated (a) shortwave radiation,(b) average air temperature, and (c) average vapor pressuredeficit. The length of the hydroperiod is denoted by thewidth of the gray bars.

Table 1. Environmental Conditions at the SJFM From 1999 to 2003

YearAir Temperature Average

(Minimum, Maximum) (�C)

SolarRadiation

(MJ m�2 a�1)

AverageVPD(kPa)

Precipitation,January–December

(mm)

HydroperiodLength (Days)(Start, Finish)

1999 15.8 (10.7, 21.1) 6909 0.53 136 87 (79, 166)2000 17.1 (12.4, 22) 6831 0.57 259 148 (18, 166)2001 16.6 (11.8, 21.5) 6445 0.54 389 126 (36, 162)2002 16.2 (10.8, 21.8) 6634 0.56 164 126 (25, 151)2003 17.1 (11.7, 22.6) 6415 0.60 382 208 (�11, 197)


4 of 12

G04019

mornings. Sensible heat fluxes decreased in the middle ofthe summer as more energy was partitioned to latent heat.Peak rates of latent heat lagged peak rates of sensible heatby 3 to 4 months (Figures 3a and 3b). High rates of sensibleheat flux occurred during the 2001 and 2002 growingseasons, and low rates during 1999, 2000 and 2003. Peakrates of sensible heat fluxes ranged from 375 W m�2 in2000 to 550 W m�2 in 2002. Although peak rates ofsensible heat were usually associated with the middle ofthe hydroperiod, the presence or absence of standing waterat the SJFM had little direct effect on the sensible heat flux.[21] Latent heat flux also demonstrated large intra and

interannual variability (Figure 3b). Latent heat peakedduring midsummer and was lowest in the winter as thecanopy senesced. Intraannual latent heat fluxes were morerelated to canopy development than to the presence orabsence of standing water. Goulden et al. [2007] demon-strated that plant transpiration was more important thanevaporation for controlling fluxes of latent heat at the SJFM,because a thick litter layer keeps water temperatures cooland rates of evaporation low. Interannual variability in latentheat fluxes was large, with peak midsummer rates of latent

heat ranging from 150 W m�2 in 2002 to 325 W m�2 in2000. Years with low latent heat flux (2001 and 2002) werecoincident with years with high sensible heat flux, asrequired by conservation of energy.[22] The NEE of CO2 also demonstrated large intra and

interannual variability (Figure 3c). The seasonal cycle ofNEE was closely correlated with the seasonal cycle of latentheat, underscoring the importance of canopy developmentin controlling both fluxes. Positive NEE occurred at nightand during the winter, whereas negative NEE occurredduring spring and summer days. There was large interan-nual variability in daytime as well as nighttime NEE. Peakrates of daytime NEE ranged from �11 mmol m�2 s�1 in2001 to �26 mmol m�2 s�1 in 2003. Peak rates of nighttimeNEE ranged from 4.5 mmol m�2 s�1 in 1999 to 8.5 mmolm�2 s�1 in 2001 and 2002. Nighttime NEE values indicatethat respiration was highest in 2001 and 2002, and lowest in1999, 2000, and 2003. Daytime NEE values indicate thatnet carbon uptake was highest in 2000 and 2003 and lowestin 2001 and 2002.[23] Interannual variability in the monthly averages of

midday Bowen ratio, daily gross carbon uptake (GEE) and

Figure 2. Canopy attributes at the SJFM from 1999 to 2003. (a) Interannual variability in stand density(gray bars; left axis) and maximum leaf area index (LAImax) (closed circles with line; right axis).(b) Interannual variability in aboveground net primary production (ANPP). Error bars represent standarderrors.


5 of 12

G04019

daily carbon loss (TER) reflected the variability observed inthe half hourly data (Figure 4). The midday Bowen ratiowas highest in the winter months (Figure 4a), indicating thatmost of the energy was dissipated as sensible heat duringthis period. There was some interannual variability in thewinter (Jan–Mar) midday Bowen ratios, but the data duringthis period had a high variance. Growing season middayBowen ratios during 2001 and 2002 were significantlyhigher than in 1999, 2000, and 2003. Average daily GEE

was highest during 2000 and 2003 (Figure 4b). Peak rates ofaverage daily TER usually lagged peak rates of photosyn-thesis by a month and were highest during the 2001 and2002 growing season (Figure 4c).[24] The seasonality of CO2 uptake was consistent from

year to year, with a peak in July (Figure 4b). The rates ofCO2 uptake did not change markedly coincident withdrying, either between or within years. The rates of CO2

uptake increased from June to July of each year, even

Figure 3. Half hourly surface energy and mass exchanges at the SJFM from 1999 to 2003. Panelsrepresent half hourly fluxes of (a) sensible heat, (b) latent heat, and (c) net ecosystem exchange of CO2

(NEE). The length of the hydroperiod is denoted by the width of the gray bars.


6 of 12

G04019

though the marsh was drying at that point. The increase inCO2 uptake from June to July was consistent from year toyear, even though the marsh was wet during this period insome years (2003) and dry in others (2002). The timing ofthe onset of CO2 uptake varied considerably from year toyear, with early starts in 2000 and 2003. However, the onsetof GEE in spring was not closely related to the presence orabsence of water, except possibly for 1999, which wasunusually dry. The onset of GEE was rapid in 2000 and

2003 and slow in 2001 and 2002, even though the marshwas flooded throughout the winter and spring of all fouryears. The presence or absence of standing water after theinitial flooding period does not appear to exert a strongdirect effect on the interannual variation of CO2 exchange inmidsummer, or on the timing of GEE development inspring.

4.4. Interannual Variability in Annual Carbon Budget

[25] Interannual variability in the annual sums of GEE,TER and NEP were large (Figure 5). NEP ranged from –251 gC m�2 a�1 in 2000 to 515 gC m�2 a�1 in 2001. Grossuptake ranged from 1090 gC m�2 a�1 in 1999 to 1639 gCm�2 a�1 in 2000, while TER ranged from 1313 gC m�2 a�1

in 1999 to 1632 gC m�2 a�1 in 2001. Larger annual GEEswere observed in 2000 and 2003 and were associated withyears that resulted in carbon sinks. Years with low GEEwere associated with a net annual carbon loss. Interannualvariability in annual TER was 73% less than the variabilityin GEE. NEP was better correlated with annual GEE (r2:0.87; p = 0.02) than TER (r2: 0.56; p = 0.14), indicating thatinterannual variability in NEP was strongly controlled byGEE.[26] The annual sums of GEE, TER, and NEE differed in

their relationship to weather and length of the hydroperiod.Interannual variability in TER was unrelated to annual airtemperature (p = 0.92), annual precipitation (p = 0.69), orthe length of the hydroperiod (p = 0.84), while GEE wasunrelated to the average potential growing season temper-ature (DOY: 100–300) (p = 0.12) and length of the hydro-period (p = 0.82), but weakly related to average annual airtemperature (r2: 0.60; p = 0.05). The effect of warmertemperatures on GEE did not directly translate to interan-nual differences in carbon storage and NEP was unrelated toannual air temperature (p = 0.21), solar radiation (p = 0.82),or length of the hydroperiod (p = 0.23).[27] The weak correlation between annual air temperature

and GEE possibly reflects the longer growing seasonbrought on by warm spring temperatures in 2000 and2003 (Figure 1), and we therefore investigated whether airtemperature helps to explain the start of the growing season.Since growing degree days are sensitive to the base tem-perature, we ran a Monte Carlo simulation that changed thebase temperature, recalculated growing degree days, andpredicted the start of the growing season based on a range ofgrowing degree days, which were constrained by actualobservations (Figure 6). The simulation ran 400 correlationsand none of the predictions came close to predicting thestart of the growing season, indicating that air temperaturealone cannot explain the start of the growing season (Figure 6,inset panel).

4.5. Interannual Variability in Vegetation Indices

[28] Midsummer EVI and midday PAR albedo (PARalb)recorded changes in surface reflectance during the growingseasons of 1999–2003 (Figure 7). The seasonality ofPARalb represented the seasonality of green leaves. Interan-nual midsummer EVI variability followed changes in eco-system physiology from 1999 to 2003 in predictable ways.EVI was low during years with reduced net and gross CO2

uptake, and increased midday Bowen ratios (1999, 2001,and 2002); EVI was high during years with increased net

Figure 4. Monthly averaged (a) midday Bowen ratio,(b) daily gross carbon uptake (GEE), and (c) dailyecosystem carbon loss (TER) from 1999 to 2003. Closedcircles represent data for 1999, open circles for 2000,closed triangles for 2001, open triangles for 2002 and closedsquares for 2003. Lines in Figure 4c represent the length ofthe hydroperiod in each year as designated by theaccompanying text on the left of each line. Error barsrepresent standard errors.


7 of 12

G04019

and gross CO2 uptake, and decreased midday Bowen ratios(2000 and 2003). Midsummer EVI was well correlated with-GEEmax (Figure 7c) (r2: 0.92; p = 0.01) and NEEmax (datanot shown) (r2: 0.89; p = 0.01) indicating that changes insurface greenness were related to interannual physiologicaldifferences at the SJFM. At the same time, LAImax was notsignificantly correlated with EVI (Figure 7b) (p = 0.62),GEEmax (p = 0.54) or NEEmax (p = 0.63), implying a changein photosynthetic efficiency between years (the rates ofgross or net CO2 exchange per LAI). Midsummer EVIwas unrelated to annual air temperature (p = 0.22), potentialgrowing season temperature (p = 0.39), and length of thehydroperiod (p = 0.09).

4.6. Long-Term Carbon Budget

[29] The long-term (1999–2003) carbon budget indicatesthat the SJFM was a source of 680 gC m�2 to theatmosphere (Figure 8). The low photosynthetic and highrespiration rates in 2001 and 2002 heavily influenced thecumulative sum of carbon and switched the SJFM from asystem that was approximately at steady state to one thatwas losing carbon (Figure 8b).

5. Discussion

[30] Interannual NEP variability at the SJFM, as indicatedby the standard deviation of NEP, was the largest that hasbeen reported for any terrestrial ecosystem (Table 2). Thelarge interannual variability in several independent meas-ures of ecosystem function (i.e., NEP, ANPP, EVI) indicatesthat the eddy covariance observations represented ‘‘true’’variability, and were not mere methodological artifact. Asurvey of the literature indicates that Duke Forest’s loblollypine plantation had the second largest variability, followedby a Southeast Asian tropical peat swamp. The differencebetween maximum and minimum NEP at the SJFM was83% larger than at the pine plantation and the standarddeviation of NEP for the SJFM was 118% larger than at thepine plantation. The large interannual NEP variability at theSJFM is surprising given that the annual temperature and

water variability at the marsh is less than that reported forother eddy covariance sites (Table 1).

5.1. Does High Plant Productivity at the SJFM Resultin a Large Carbon Sink?

[31] The SJFM was a source of carbon over five yearsdespite high rates of aboveground NPP, which were typicalof other freshwater marshes (Figures 2 and 8). The towermeasurements may have underestimated this source some-what, since additional carbon may have been exported asDissolved Organic Carbon (DOC) or Particulate OrganicCarbon (POC). However, these additional exports werelikely minor, since most water is lost from the marshthrough evapotranspiration and subsurface flow, and sincethe closed basin hydrology of the SJFM prevents the lateral

Figure 5. Annual rates of ecosystem respiration [TER] (black bar), gross carbon uptake [-GEE] (lightgray bar), and the annual carbon balance [NEP] (dark gray bar) from 1999 to 2003.

Figure 6. The relationship between start of the growingseason and growing degree days with a base temperatureequal to 10�C. The inset plot includes a histogram ofPearson correlation coefficients of modeled and predictedstart of the growing season with different base temperaturesand growing degree days. The right side of the hatched linerepresents correlations that would be significant at the 95%confidence level.


8 of 12

G04019

movement of large pieces of organic material. For example,if we assume the annual subsurface and lateral outflow was50 cm and the DOC concentration was equal to themaximum concentration reported for a constructed wetlandin Florida [Stern et al., 2007], the DOC export was �10 gCm�2 a�1, implying that Figure 8 understates the loss ofcarbon by �10%[32] The inability of the SJFM to sequester carbon

resulted from consistently high respiration rates, and two

years with lower than average carbon uptake (i.e., 2001 and2002). Aerial photographs and personal communicationindicate the marsh is approximately fifty years old [Gouldenet al., 2007]. It is likely that the ability of seasonally floodedfreshwater marshes to sequester carbon diminishes with agedue to the accumulation of organic matter and the stimula-tion of ecosystem respiration by increased substrate. Previ-ous researchers have suggested that the high productivity offreshwater marshes make them ideal ecosystems for carbon

Figure 7. (a) Midsummer EVI (closed dark circles), 1-midday PAR albedo (closed gray circles) and therelationship between (b) EVI and LAImax and (c) EVI and –GEEmax for the SJFM from 1999 to 2003.Numbers in inset plots indicate respective years.

Figure 8. (a) Long-term and (b) annual carbon budget for the SJFM from 1999 to 2003. The length ofthe hydroperiod is denoted by the width of the gray bars in the main plot, while numbers beside Figure 8bdesignate representative years of annual cumulative sums.


9 of 12

G04019

sequestration [Thom et al., 2002]. However, our resultsemphasize the need to understand the individual compo-nents of an ecosystem’s carbon balance in order to assess itscarbon sequestration potential.

5.2. Physiological Characteristics of the SJFM andTheir Relationship to Environmental Factors

[33] The SJFM is unusual in that much of the interannualvariability in NEP was attributable to changes in GEE,whereas the attribution of NEP variability in other ecosys-tems has focusedmore on ecosystem respiration [cf.Valentiniet al., 2000; Barford et al., 2001;Hollinger et al., 2004;Dunnet al., 2007]. This indicates that photosynthesis is the key tointerpreting SJFM interannual ecosystem function variabil-ity. High leaf areas did not translate into high photosyntheticrates in 2001 and 2002, resulting in the poor relationshipsbetween LAImax and both EVI and the peak rates ofphotosynthesis (Figure 7). This finding is important becauseit diverges from previous work that has established LAI asthe main factor that connects vegetation indices and canopygas exchange [Gamon et al., 1995; Rahman et al., 2005].Our findings imply that changes in the photosyntheticefficiency of the canopy (the rates of gross or net CO2

exchange per LAI) are the main cause of interannualecosystem function variability at the SJFM.

5.3. Is the Physical Environment Directly Responsiblefor Interannual NEP Variability at the SJFM?

[34] Previous work has attributed interannual GEE, TER,and NEP variability to the direct effects of physical factorssuch as temperature and precipitation on ecosystem physi-

ology [cf. Barford et al., 2001; Flanagan et al., 2002; Dunnet al., 2007]. Ecosystems with the highest interannualvariability in temperature or water availability might there-fore be expected to have the highest interannual NEPvariability. However, we found that the interannual NEPvariability at the SJFM was the largest that has beenreported for any terrestrial ecosystem, even though thevariability in average annual temperature was equal to orless than that observed for many ecosystems (Tables 1 and2). This finding calls into question the idea of a universal,tight link between weather and interannual GEE, TER, andNEP variability.[35] The effects of interannual weather variability on

ecosystem physiology would be expected to play out in acouple of ways: (1) through strong, direct correlationsbetween specific weather conditions and CO2 exchange(for example, a decrease in CO2 uptake coincident withdrying soil or unusually warm temperatures), or (2) througha chain of strong correlations between specific weatherconditions, the development or loss of LAI, and the effectof LAI on CO2 exchange (for example, the rapid develop-ment of LAI with spring warming, and a subsequentincrease in CO2 uptake). In fact, we were unable to findclear evidence that either of these factors plays an importantrole at the SJFM.[36] Many processes at the SJFM were uncorrelated with

the immediate physical conditions (Figures 2, 5, and 7 andTable 1). Differences in the start of the growing season werenot well correlated with either the presence of standingwater (Figure 4) or temperature (Figure 6). The seasonal

Table 2. Comparison of Interannual NEP Variability Among Different Ecosystemsa

Ecosystem TypeNumberof Years

NEPmax- NEPmin

(gC m�2)

StandardDeviationof NEP Source

Evergreen ForestsLoblolly Pine 7 418 166 Oren et al. [2006]Boreal Evergreen 7 140 45 Hollinger et al. [2004]Boreal Evergreen 10 106 30 Dunn et al. [2007]Sub-alpine forest 2 23 16 Monson et al. [2002]Boreal Evergreen 3 19 10 Arain et al. [2002]

Deciduous ForestsBoreal Aspen 5 210 81 Arain et al. [2002]Temperate Deciduous 9 130 38 Barford et al. [2001]Northern Hardwood 5 97 41 Gough et al. [2008]Temperate Deciduous 3 50 28 Ehman et al. [2002]Temperate Deciduous 5 59 27 Wilson and Baldocchi [2001]

GrasslandsTallgrass Prairie 3 228 116 Suyker et al. [2003]Annual Grassland 2 161 114 Xu and Baldocchi [2004]Temperate Grassland 2 39 27 Flanagan et al. [2002]

Tropical ForestAmazonian Forest 3 45 24 Saleska et al. [2003]

DesertsDesert Shrub 2 13 9 Hastings et al. [2005]

WetlandsFreshwater Marsh 5 766 363 This StudySub-arctic Sedge Fen 5 311 134 Griffis and Rouse [2001]Tropical Peat Swamp 3 289 151 Hirano et al. [2007]Ombrotrophic bog 4 241 113 Lafleur et al. [2003]Alpine Meadow 3 114 63 Kato et al. [2006]

aThe SJFM is in bold, and other values were compiled from the literature.


10 of 12

G04019

patterns of GEE, TER, and midday Bowen ratio wereindependent of the length of the hydroperiod and the initialtiming of flooding. Air temperature was weakly correlatedwith annual GEE, but the unrealistic sensitivity of annualGEE to temperature indicates that this relationship wasmore coincident than causative. Attributing interannualGEE variability to temperature would require a temperaturesensitivity that is nine times greater than that reportedpreviously [Law et al., 2002]. Finally, LAImax and land-atmosphere exchange were largely uncorrelated, implyingthat the causal relationships between weather, LAI devel-opment, and exchange were weak.[37] In conclusion, the interannual variation in carbon

exchange at the SJFM resulted from shifts in the marsh’sproduction efficiency that were not caused by changes in theweather. Our findings challenge the assumption that inter-annual variation of land-atmosphere exchange is universallycaused by the direct effect of weather on ecosystem phys-iology, and that an ecosystem’s physiological response tothe physical environment is consistent from year-to-year.Future work will focus on further understanding the mech-anisms behind interannual variability at the SJFM [Rocha etal., 2008; Rocha and Goulden, 2008].

[38] Acknowledgments. This work was supported by the Universityof California, Irvine, and by the University of California Water ResourcesCenter. This work was performed at the University of California NaturalReserve System San Joaquin Freshwater Marsh Reserve. We thank BillBretz and Peter Bowler for managing and supporting the San JoaquinMarsh.

ReferencesArain, M. A., T. A. Black, A. G. Barr, P. G. Jarvis, J. M. Massheder, D. L.Verseghy, and Z. Nesic (2002), Effects of seasonal and interannual cli-mate variability on net ecosystem productivity of boreal deciduous andconifer forests, Can. J. For. Res., 32, 878–891, doi:10.1139/x01-228.

Barford, C. C., S. C. Wofsy, M. L. Goulden, J. W. Munger, E. H. Pyle, S. P.Urbanski, L. Hutyra, S. R. Saleska, D. Fitzjarrald, and K. Moore (2001),Factors controlling long- and short-term sequestration of atmosphericCO2 in a mid-latitude forest, Science, 294, 1688–1691, doi:10.1126/science.1062962.

Bradbury, I. K., and J. Grace (1983), Primary production in wetlands, inEcosystems of the World, vol. 4A, edited by A. J. P. Gore, 440 pp.,Elsevier, Amsterdam.

Chander, G., and B. Markham (2003), Revised Landsat-5 TM radiometriccalibration procedures and post-calibration dynamic ranges, IEEE Trans.Geosci. Remote Sens., 41, 2674–2677, doi:10.1109/TGRS.2003.818464.

Chen, W. J., et al. (1999), Effects of climatic variability on the annualcarbon sequestration by a boreal aspen forest, Global Change Biol., 5,41–53, doi:10.1046/j.1365-2486.1998.00201.x.

Davidson, E. A., I. A. Janssens, and Y. Luo (2006), On the variability ofrespiration in terrestrial ecosystems: Moving beyond Q10, GlobalChange Biol., 12, 154–164, doi:10.1111/j.1365-2486.2005.01065.x.

Dunn, A. L., C. C. Barford, S. C. Wofsy, M. L. Goulden, and B. C. Daube(2007), A long-term record of carbon exchange in a boreal black spruceforest: Means, responses to interannual variability, and long-term trends,Global Change Biol., 13, 577–590, doi:10.1111/j.1365-2486.2006.01221.x.

Ehman, J. L., H. P. Schmid, C. S. B. Grimmond, J. C. Randolph, P. J.Hanson, C. A. Wayson, and F. D. Cropley (2002), An initial intercom-parison of micrometeorological and ecological inventory estimates ofcarbon exchange in a mid-latitude deciduous forest, Global Change Biol.,8, 575–589, doi:10.1046/j.1365-2486.2002.00492.x.

Esser, G., H. F. H. Lieth, J. M. Scurlock, and R. J. Olson (2000), Osnabrucknet primary productivity data set, Ecology, 81, 1177.

Flanagan, L. B., L. A. Wever, and P. J. Carlson (2002), Seasonal andinterannual variation in carbon dioxide exchange and carbon balance ina northern temperate grassland, Global Change Biol., 8, 599 –615,doi:10.1046/j.1365-2486.2002.00491.x.

Gamon, J. A., C. B. Field, M. L. Goulden, K. L. Griffin, A. E. Hartley,G. Joel, J. Penuelas, and R. Valentini (1995), Relationship betweenNDVI, canopy structure, and photosynthesis in three Californian vegeta-tion types, Ecol. Appl., 5, 28–41, doi:10.2307/1942049.

Gough, C. M., C. S. Vogel, H. P. Schmid, H.-B. Su, and P. S. Curtis (2008),Multi-year convergence of biometric and meteorological estimates offorest carbon storage, Agric. For. Meteorol. , 148, 158 – 170,doi:10.1016/j.agrformet.2007.08.004.

Goulden, M. L., J. W. Munger, S. M. Fan, B. C. Daube, and S. C. Wofsy(1996), Exchange of carbon dioxide by a deciduous forest: Response tointerannual climate variability, Science, 271, 1576–1578, doi:10.1126/science.271.5255.1576.

Goulden, M. L., B. C. Daube, S.-M. Fan, D. J. Sutton, A. Bazzzaz, J. W.Munger, and S. C. Wofsy (1997), Physiological Responses of a BlackSpruce Forest to Weather, J. Geophys. Res., 102, 28,987 – 28,996,doi:10.1029/97JD01111.

Goulden, M. L., M. C. Litvak, and S. D. Miller (2007), Factors controllingmarsh evapotranspiration, Aquat. Bot., 58, 97–106, doi:10.1016/j.aqua-bot.2006.09.005.

Griffis, T. J., and W. R. Rouse (2001), Modeling the interannual variabilityof net ecosystem CO2 exchange at a subarctic sedge fen, Global ChangeBiol., 7, 511–530, doi:10.1046/j.1354-1013.2001.00403.x.

Hall, F. G., D. E. Strebel, J. E. Nickeson, and S. J. Goetz (1991), Radio-metric rectification-toward a common radiometric response among multi-date, multisensor images, Remote Sens. Environ., 35, 11 – 27,doi:10.1016/0034-4257(91)90062-B.

Hanson, P. J., et al. (2004), Oak forest carbon and water simulations:Monthly intercomparisons and evaluations against independent data,Ecol. Monogr., 74, 443–489, doi:10.1890/03-4049.

Hastings, S. J., W. C. Oechel, and A. Muhlia-Melo (2005), Diurnal, seaso-nal and annual variation in the net ecosystem CO2 exchange of a desertshrub community (Sarcocaulescent) in Baja California, Mexico, GlobalChange Biol., 11, 927–939, doi:10.1111/j.1365-2486.2005.00951.x.

Hirano, T., H. Segah, T. Harada, S. Limin, T. June, R. Hirata, and M. Osaki(2007), Carbon dioxide balance of a tropical peat swamp forest in Kali-mantan, Indonesia, Global Change Biol., 13, 412–425, doi:10.1111/j.1365-2486.2006.01301.x.

Hollinger, D. Y., et al. (2004), Spatial and temporal variability in forest-atmosphere CO2 exchange, Global Change Biol., 10, 1689–1706,doi:10.1111/j.1365-2486.2004.00847.x.

Huete, A. T., K. Didan, T. Miura, E. P. Rodriguez, X. Gao, and L. G.Ferreira (2002), Overview of the radiometric and biophysical perfor-mance of the MODIS vegetation indices, Remote Sens. Environ., 83,195–213, doi:10.1016/S0034-4257(02)00096-2.

Hui, D., Y. Luo, and G. Katul (2003), Partitioning interannual variability innet ecosystem exchange between climatic variability and functionalchange, Tree Physiol., 23, 433–442.

Joiner, D. W., P. M. Lafleur, J. H. McCaughey, and P. A. Bartlett (1999),Interannual variability in carbon dioxide exchanges at a boreal wetlandsin the BOREAS northern study area, J. Geophys. Res., 104, 27,663–27,672, doi:10.1029/1999JD900136.

Kato, T., Y. Tang, S. Gu, M. Hirota, M. Du, L. Yingnian, and Z. Xinquan(2006), Temperature and biomass influences on interannual changes inCO2 exchange in an alpine meadow on the Qinghai-Tibetan Plateau, GlobalChange Biol., 12, 1285–1298, doi:10.1111/j.1365-2486.2006.01153.x.

Kelly, R. H., W. J. Parton, M. D. Hartman, L. K. Stretch, D. S. Ojima, andD. S. Schimel (2000), Intra-annual and interannual variability of ecosys-tem processes in shortgrass steppe, J. Geophys. Res., 105, 20,093–20,100, doi:10.1029/2000JD900259.

Lafleur, P. M., N. T. Roulet, J. L. Bubier, S. Frolking, and T. R. Moore(2003), Interannual variability in the peatland-atmosphere carbon dioxideexchange at an ombrotrophic bog, Global Biogeochem. Cycles, 17(2),1036, doi:10.1029/2002GB001983.

Law, B. E., et al. (2002), Environmental controls over carbon dioxide andwater vapor exchange of terrestrial vegetation, Agric. For. Meteorol., 113,97–120, doi:10.1016/S0168-1923(02)00104-1.

LeBauer, D. S., and K. K. Treseder (2008), Nitrogen limitation of netprimary productivity in terrestrial ecosystems is globally distributed,Ecology, 89, 371–379, doi:10.1890/06-2057.1.

Monson, R. K., A. A. Turnipseed, and J. P. Sparks (2002), Carbon seques-tration in a high-elevation, subalpine forest, Global Change Biol., 8,459–478, doi:10.1046/j.1365-2486.2002.00480.x.

Oren, R., C.-I. Hsieh, P. Stoy, J. Albertson, H. R. McCarthy, P. Harrell, andG. G. Katul (2006), Estimating the uncertainty in annual net ecosystemcarbon exchange: Spatial variation in turbulent fluxes and sampling errorsin eddy-covariance measurements, Global Change Biol., 12, 883–896,doi:10.1111/j.1365-2486.2006.01131.x.

Rahman, A. F., D. A. Sims, V. D. Cordova, and B. Z. El-Masri (2005),Potential of MODIS EVI and surface temperature for directly estimatingper-pixel ecosystem C fluxes, Geophys. Res. Lett., 32, L19404,doi:10.1029/2005GL024127.

Richardson, A. D., et al. (2006), Comparing simple respiration models foreddy flux and dynamic chamber data, Agric. For. Meteorol., 141, 219–234, doi:10.1016/j.agrformet.2006.10.010.


11 of 12

G04019

Richardson, A. D., A. Y. Hollinger, J. D. Aber, S. V. Ollinger, and B. H.Braswell (2007), Environmental variation is directly responsible forshort- but not long-term variation in forest-atmosphere carbon exchange,Global Change Biol., 13, 788–803.

Rocha, A. V., andM. L. Goulden (2008),Why is marsh productivity so high?New insights from eddy covariance and biomass measurements in a Typhamarsh, Agric. For. Meteorol., 113, G04020, doi:10.1016/j.agrformet.2008.07.010.

Rocha, A. V., D. L. Potts, and M. L. Goulden (2008), Standing litter as adriver of interannual CO2 exchange variability in a freshwater marsh, J.Geophys. Res., doi:10.1029/2008JG000713, in press.

Ruimy, A., P. G. Jarvis, D. D. Baldocchi, and B. Saugier (1995), CO2 fluxesover plant canopies: A literature review, Adv. Ecol. Res, 26, 1 –68,doi:10.1016/S0065-2504(08)60063-X.

Saleska, S. R., et al. (2003), Carbon in Amazon forests: Unexpected sea-sonal fluxes and disturbance-induced losses, Science, 302, 1554–1557,doi:10.1126/science.1091165.

Stern, J., Y. Want, B. Gu, and J. Newman (2007), Distribution and turnoverof carbon in natural and constructed wetlands in the Florida Everglades,Appl. Geochem., 22, 1936–1948, doi:10.1016/j.apgeochem.2007.04.007.

Suyker, A. E., S. B. Verma, and G. C. Burba (2003), Interannual variabilityin net CO2 exchange of a native tallgrass prairie, Global Change Biol., 9,255–265, doi:10.1046/j.1365-2486.2003.00567.x.

Thom, R. M., S. L. Blanton, A. B. Borde, G. D. Williams, D. L. Woodruff,M. H. Huesemann, K. W. Nehring, and S. E. Brauning (2002), Investiga-tions into wetland carbon sequestration as remediation for global warming,in Wetlands & Remediation II: Proceedings of the Second InternationalConference on Wetlands & Remediation, Salt Lake City, UT, USA, 16–17November 1999, pp. 311–320, Battelle, Columbus, Ohio.

Thornton, P. E., et al. (2002), Modeling and measuring the effects of dis-turbance history and climate on carbon and water budgets in evergreenneedleleaf forest, Agric. For. Meteorol., 113, 185–222, doi:10.1016/S0168-1923(02)00108-9.

Twine, T. E., W. P. Kustas, J. M. Norman, D. R. Cook, P. R. Houser, T. P.Meyers, J. H. Prueger, P. J. Starks, and M. L. Wesely (2000), Correctingeddy-covariance flux underestimates over a grassland, Agric. For. Me-teorol., 103, 279–300, doi:10.1016/S0168-1923(00)00123-4.

Urbanski, S., C. Barford, S. Wofsy, C. Kucharik, E. Pyle, J. Budney,K. Mckain, D. Fitzjarrald, M. Czikowsky, and J. W. Munger (2007),Factors controlling CO2 exchange on timescales from hourly to decadalat Harvard Forest, J. Geophys. Res., 112, G02020, doi:10.1029/2006JG000293.

Valentini, R., et al. (2000), Respiration as the main determinant of carbonbalance in European forests, Nature, 404, 861 – 865, doi:10.1038/35009084.

Wilson, K. B., and D. D. Baldocchi (2001), Comparing independent esti-mates of carbon dioxide exchange over 5 years at a deciduous forest inthe southeastern United States, J. Geophys. Res., 106, 34,167–34,178,doi:10.1029/2001JD000624.

Xu, L., and D. D. Baldocchi (2004), Seasonal variation in carbon dioxideexchange over a Mediterranean annual grassland in California, Agric.For. Meteorol., 123, 79–96, doi:10.1016/j.agrformet.2003.10.004.

��M. L. Goulden, Department of Earth System Science, University of

California, Irvine, CA 92697, USA.A. V. Rocha, Marine Biological Laboratory, 7 MBL Street, Woods Hole,

MA 02345, USA. ([email protected])


12 of 12

G04019

Large interannual CO2 and energy exchange variability in a freshwater marsh under consistent...

Documents

Transcript of Large interannual CO2 and energy exchange variability in a freshwater marsh under consistent...